CN101896690B

CN101896690B - Parallel adaptive data partitioning on a reservoir simulation using an unstructured grid

Info

Publication number: CN101896690B
Application number: CN200880120763.2A
Authority: CN
Inventors: A·K·乌萨达; I·米舍夫
Original assignee: Exxon Production Research Co
Current assignee: ExxonMobil Upstream Research Co
Priority date: 2007-12-13
Filing date: 2008-10-20
Publication date: 2015-02-18
Anticipated expiration: 2028-10-20
Also published as: BRPI0820870A2; EP2247820A4; US20130246030A1; EP2247820A1; CN101896690A; CA2702965C; US8725481B2; US20100217574A1; US8437996B2; CA2702965A1; WO2009075945A1

Abstract

A computer implemented system and method for parallel adaptive data partitioning on a reservoir simulation using an unstructured grid includes a method of simulating a reservoir model which includes generating the reservoir model. The generated reservoir model is partitioned into multiple sets of different domains, each one corresponding to an efficient partition for a specific portion of the model.

Description

Use the parallel adaptive data partitioning on the reservoir simulation of unstructured grid

The cross reference of related application

This application claims the rights and interests of the U.S. Provisional Patent Application 61/007470 of application on December 13rd, 2007, its title is " PARALLEL ADAPTIVE DATA PARTITIONING ONRESERVOIR SIMULATION USING AN UNSTRUCTURE GRID ", and the full content of this application is included in this by reference.

Technical field

The present invention relates generally to production of hydrocarbons, and be specifically related to use reservoir simulation to promote production of hydrocarbons.

Background technology

Summary of the invention

In one aspect, the method simulating reservoir model comprises generation stored models; Be the set of multiple not same area by the reservoir model subregion of generation, eachly correspond to the efficient zoned of specific part model.

The enforcement of this aspect can comprise the one or more of lower region feature.Such as, simulate reservoir model can comprise divide reservoir be modeled as multiple processing element; And based on the multiple processing element of Paralleled process.Parallel Simulation reservoir simulation can comprise and dynamically rezoning the reservoir model of generation for multiple territory is to improve parallel performance.The reservoir model of generation is rezoned for multiple territory can comprise a) by selecting partition scheme and determining its parameter pretreatment reservoir model; B) be multiple territory with this partition scheme by the reservoir model subregion of generation; C) post processing subregion reservoir model thus improve further subregion calculate parallel performance; D) quality of the subregion reservoir model of post processing is assessed; If with the quality of e) the subregion reservoir model of post processing is lower than predetermined value, then the partition scheme of correction and parameter is used to repeat a, b, c, d and e.Be the Node subsets or block that multiple territory can comprise that identification isolates each other by the reservoir model subregion of generation; The joint block that weighting is selected thus the processing cost that associates with each piece is described; The joint block of these joint blocks and weights assigned is selected to corresponding territory based on processing cost.Be that multiple territory can comprise the level determining the processing cost associated with each node in the reservoir model generated by the reservoir model subregion of generation; Chosen node on geometric direction; Based on processing cost storage weighting chosen node thus generate equal weight storehouse (bin); With from storehouse distribution node to territory.Be that multiple territory can comprise the velocity field that associates of reservoir model determined and generate by the reservoir model subregion of generation; Follow the tracks of the streamline associated with this velocity field; Projection streamline thus generate stream curtain (stream curtain); And extend stream curtain to generation reservoir model border thus be territory by the reservoir model subregion of generation.It is the processing cost that multiple territory can comprise each node of reservoir model of determining and generate and associates by the reservoir model subregion of generation; Determine the processing cost that the connection level (connectivitylevel) between each node of the reservoir model generated associates; Be multiple territory according to the processing cost determined by the reservoir model subregion of generation.

Be that multiple territory can comprise connective node more than predeterminated level in grouping same domain according to the processing cost determined with being communicated with the reservoir model subregion of level by generation.Be that multiple territory can comprise territory subregion by the reservoir model subregion of generation; Determine the distance between the interior adjacent well defined of the reservoir model of border, territory and generation; With based on the distance determined, as requested the reservoir model of generation is rezoned, territory subregion is moved away from well and therefore improves solver performance.

Territory subregion can be comprised subset or the block of the node that identification is isolated each other; These joint blocks of weighting are to illustrate the processing cost associated with each piece; The joint block of these joint blocks and weights assigned is selected to corresponding territory based on processing cost.Territory subregion can be comprised and determine and the processing cost level that each node in the reservoir model generated associates; Chosen node on geometric direction; Based on processing cost storage weighting chosen node thus generate equal weight storehouse; With from storehouse distribution node to territory.Territory subregion can be comprised and determine and the velocity field that the reservoir model generated associates; Follow the tracks of the streamline associated with velocity field; Projection streamline thus generate stream curtain; And extend stream curtain to generation reservoir model border thus be territory by the reservoir model subregion of generation.Territory subregion can be comprised and determine and the processing cost that each node of the reservoir model generated associates; Determine each internodal processing cost being communicated with horizontal relevance with the reservoir model generated; Be multiple territory according to the processing cost determined by the reservoir model subregion of generation.Be that multiple territory can comprise territory subregion by the reservoir model subregion of generation; Determine all nodes along boundary alignment between territory in the reservoir model that generates; Projected boundary node to plane and matching by the curve of boundary node of projection; And in the direction Projective Curve orthogonal with matched curve thus between the territory defining the reservoir model of generation again border.

Be that multiple territory can comprise the performance compared the parallel performance of reservoir model subregion of generation and the history set of the reservoir model of subregion by the reservoir model subregion of generation; If with the performance of new subregion and the performance of historical record different good, then model is rezoned.

In yet another aspect, the method simulating reservoir model comprises generation reservoir model; Be multiple territory by the reservoir model subregion of generation; What divide reservoir is modeled as multiple processing element; The multiple processing element of parallel processing; With in parallel process, be other multiple territory by the reservoir model subregion of generation at least one times; Be wherein that multiple territory comprises by the reservoir model subregion of generation: a) by selecting partition scheme and determining its parameter pretreatment reservoir model; B) be multiple territory with partition scheme by the reservoir model subregion of generation; C) post processing subregion reservoir model thus correct the reservoir model of subregion, improve the parallel performance that subregion calculates further; D) quality of the subregion reservoir model of post processing is assessed; And if the quality of e) the subregion reservoir model of post processing is lower than predetermined value, then repeat a, b, c, d with the partition scheme suitably revised and/or its parameter, and e.

In another is general, the method for simulation reservoir model can comprise generation reservoir model; Be multiple territory by the reservoir model subregion of generation; What divide reservoir is modeled as multiple processing element; The multiple processing element of parallel processing; Be other multiple territory by the reservoir model subregion of generation at least one times in parallel process.

It is any step that multiple territory can comprise in step below by the reservoir model subregion of generation.Particularly, territory subregion can be comprised subset or the block of the node that identification is isolated each other; The joint block that weighting is selected thus the processing cost that associates with each piece is described; These joint blocks are selected based on processing cost; Corresponding territory is given with the joint block of weights assigned.Territory subregion can be comprised and determine and the processing cost level that each node in the reservoir model generated associates; Chosen node on geometric direction; Based on processing cost storage weighting chosen node thus generate the storehouse of equal weight; From storehouse distribution node to territory.Territory subregion can be comprised and determine and the velocity field that the reservoir model generated associates; Follow the tracks of the streamline associated with velocity field; Projection streamline thus generate stream curtain; With extend stream curtain to generation reservoir model border thus be territory by the reservoir model subregion of generation.Territory subregion can be comprised and determine and the processing cost that each node of the reservoir model generated associates; The processing cost of the connective horizontal relevance between each node of the reservoir model determined and generate; Be multiple territory according to the processing cost determined by the reservoir model subregion of generation.Be that multiple territory can comprise territory subregion by the reservoir model subregion of generation; Determine all nodes along boundary alignment between territory in the reservoir model that generates; Projected boundary node to plane and matching by the curve of boundary node of projection; And in the direction Projective Curve orthogonal with matched curve thus between the territory defining the reservoir model of generation again border.

One or more aforementioned aspect can be used to simulate reservoir model, and then can be dependent on this, and the analog result according to reservoir model controls production of hydrocarbons activity.The production of hydrocarbon can be controlled, the output control that the productivity ratio as earth's surface facility can be understood based on (multiple) reservoir model from simulation.

Accompanying drawing explanation

Fig. 1 is the diagram of the reservoir simulation model comprising the open work mesh defining multiple node.

Fig. 2 is the flow chart of the simulator of the operation of model in simulation drawing 1.

Fig. 3 is the diagram of the reservoir simulation model comprising the open work mesh defining multiple node, and this model has divided multiple territory into.

Fig. 4 is the diagram of the reservoir simulation model comprising the open work mesh defining multiple node, and wherein in grid, different node is with the implicity of varying level and different fluid modeling.And the node of reservoir simulation model has divided two territories (0 and 1) into.

Fig. 5 is the diagram of the numerical matrix corresponding to Fig. 4 model.

Fig. 6 a is the flow chart of the simulator of the operation of simulation drawing 1 model.

Fig. 6 b is the flow chart of the well management zoned logic of simulator in Fig. 6 a.

Fig. 6 c is the flow chart of the zoned logic that constructs of the Jacobi (Jacobian) of Fig. 6 a simulator and flow rate calculation.

Fig. 6 d is the flow chart of the zoned logic of the linear solution of simulator in Fig. 6 a.

Fig. 6 e is the zoned logic flow chart of the property calculation of simulator in Fig. 6 a.

Fig. 7 is by the flow chart of the conventional method of arbitrary calculating section subregion of formation reservoir simulation process.

Fig. 8 is by the flow chart of the node-coloring method of reservoir model subregion.

Fig. 8 a to 8d is the schematic diagram of the various operating procedures of Fig. 8 interior joint colorize method.

Fig. 9 is by the flow chart of the method for geometry of the balancing the load of reservoir model subregion.

Fig. 9 a to Fig. 9 e is the schematic diagram of the various operating procedures of the method for geometry of balancing the load in Fig. 9.

Figure 10 is by the flow chart of the streamline method of reservoir model subregion.

Figure 10 a to 10c is the schematic diagram of the various operating procedures of streamline method in Figure 10.

Figure 11 is by the flow chart to oil well distance (distance-to-well) method of reservoir model subregion.

Figure 12 is the diagram of the node weighting of arriving the reservoir model of the function of oil well distance as it.

Figure 13 is by the method flow diagram of reservoir model subregion.

Figure 14 is the flow chart of the curve-fitting method of level and smooth reservoir model subregion.

Figure 14 a to 14d is the schematic diagram of the various operating procedures of the curve smoothing method of Figure 14.

Figure 15 is by the flow chart of the historical comparison approach of reservoir model subregion.

Figure 16 is the assessment diagram of the individual domain performance of reservoir model simulation neutral line solution.

Detailed description of the invention

Initial reference Fig. 1, the exemplary embodiment that the typical case 3 of the operation of simulated oil and/or gas reservoir ties up reservoir model 100 comprises one or more peupendicular hole 102.In the exemplary embodiment, model 100 is divided into multiple node 104 by open work mesh 106.In the exemplary embodiment, the node 104 of model 100 has inconsistent size.

In the exemplary embodiment, as shown in Figure 2, the operation of model 100 is simulated with conventional reservoir simulator 200, is wherein that well and the earth's surface facility network of model performs well management 202.In the exemplary embodiment, well management 202 performs all wells, as shown in model 100 102, comprises conventional iterative process 204, wherein conventional Jacobi's structure and flow rate calculation 206 are performed, and are then that conventional linear solves 208 and General Properties calculating 210.In the exemplary embodiment, linear solution 208 and/or property calculation 210 perform in large data array, and data array represents attribute, as pressure and the composition at mesh point place in grid 106.

In the exemplary embodiment, after the process 204 of well 102 completes in for model, the analogue data of whole reservoir model generates at conventional result/checkpoint I/O (I/O) 212 place.

In the exemplary embodiment, reservoir simulator 200 can utilize, and such as one or more all-purpose computer, special-purpose computer, analog processor, digital processing unit, CPU and/or distributed computing system realize.

In the exemplary embodiment, model 100 and simulator 200 are used to the operation of simulation reservoir thus allow the fluid, energy and/or the gas modeling that flow in hydrocarbon reservoir, well and relevant earth's surface facility.Reservoir simulation is a part for reservoir modeling, and reservoir modeling also comprises constructing analog data thus accurate representation reservoir.The object of simulation understands flow pattern thus optimizes certain strategy producing hydrocarbon from some well and earth's surface facility set.Simulate a part for iterative process normally consuming time, to reduce the uncertainty about particular reservoir specification of a model, optimize production strategy simultaneously.Reservoir simulation is such as the simulation of a kind of computational fluid dynamics.

The usual major part of calculating that simulator 200 performs all performs large data array, and data array represents physical attribute, as the pressure of mesh point and composition in grid 106.Along with time stepping method, the relative cost of the part operation of simulator 200 can change.Such as, linear solution 208 can become and construct the many of 206 costlinesses than Jacobi.This may be the attribute due to the physical process be modeled, or due to the attribute of algorithm.Such as, reservoir simulator 200 can from single hydrocarbon mutually.But produce due to oil along with reservoir pressure and decline, pressure can drop to below fluid bubble point, and therefore gas can be overflowed from solution.And then this can make property calculation 210 more expensive, but can not significantly affect linear solution 208.Clean effect is the total computing time making property calculation take larger percentage.And the cost of property calculation changes by grid node 104.Namely, than another region, a region of reservoir model 100 can require that more calculating is to converge on suitable solution.

In the exemplary embodiment, in order to reduce simulator 200 operation required by running time, the one or more operating procedures 202,204,206,208,210 and/or 212 of simulator can distribute, so that executed in parallel operating procedure in multiple CPU (CPU) in computer or CPU core.In the exemplary embodiment, the parallel method of the operating procedure 202,204,206,208,210 and/or 212 of simulator 200 changes by classification.Such as, can be different from the parallel method of another specific operation step of simulator by the method for the specific operation step parallelization of simulator 200.In the exemplary embodiment, the method for the parallelization selected by the specific operation step of simulator 200 can be optimized by empirical method.

In the exemplary embodiment, for the specific operation step of simulator 200, or the specific parallel method of operating procedure group selection considers and operating procedure, or whether the calculating of operating procedure group association is local, wherein a small amount of or there is no inter-domain communication, or be overall, wherein require the communication on cross-domain border.Such as, the parallelization of simulator 200, by being multiple territory and providing by model 100 subregion, in the exemplary embodiment, being optimized the balancing the load that parallelization provides and is also made the communication between model domain minimum.

In the exemplary embodiment, parallelization is by providing by tasks in parallel.In the exemplary embodiment, be by being divided into multiple subtask to provide the operating procedure of simulator 200 by tasks in parallel, subtask can parallel running therefore by multiple stage computer disposal.Such as, all or part property calculation 210 can fall into this classification, because much calculating only relates to calculating at node but not from the flow of connected node.Therefore, these calculating can simultaneously executed in parallel, and partial result nothing but.

In the exemplary embodiment, parallelization provides by the parallelization of data partition.

In the exemplary embodiment, as shown in Figure 3, the parallelization of data partition is such as, by being independent domains, 100a and 100b by the data partition in the grid 106 of model 100, and each territory of data is performed to same operating procedure provides.Such as, Jacobi construct 206 and property calculation 210 usually fall into this classification.This parallel method is normally suitable for local calculation.

In the exemplary embodiment, as shown in Figure 3, the parallelization of data partition is by the data in the grid 106 of model 100 are divided into independent domains, such as 100a and 100b, and perform parallel algorithm and provide, so that the major part of the calculating of the one or more operating procedures of simulator 200 can perform on an equal basis on the not same area of model.In the exemplary embodiment, as linear solution 208, the extra global portions calculated can be required.

In the exemplary embodiment, the calculating performed in the operating procedure of simulator 200 is by by data partition parallelization.In the exemplary embodiment, the one or more calculating of the one or more operating procedures of simulator 200 can comprise the corresponding subregion of the data of model 100.And for the one or more calculating of the one or more operating procedures of simulator 200, the optimum partition of the data of model 100 can depend on the time.Such as, in the operating process of simulator 200, in different time points, parallelization can have diverse data partition.

The existing partitioning algorithm of simulator 200 is attempted for each territory of model 100 provides effective balancing the load, and minimizes the linking number between subdomain.The method must not provide the good iteration performance of the Parallel solver decomposed based on territory.And this is the mainspring of method described in exploitation this patent.

Due to the evolution property of reservoir simulator 200, the existing subregion of model 100 may become balancing the load or inadequately for current computing mode poor efficiency.This may occur, because the cost of property calculation depends on fluid properties and can move with fluid and develop and sharply change.Or linear solution 208 can encounter the overall situation (entirety) convergence difficulties, this is because the characteristic scalar (character) of linear matrix equation changes.In such circumstances, expect the data of model 100 again subregion to make the operation of simulator 200 get back to suitable balancing the load and to improve the iteration convergence of linear solution 208.

In the exemplary embodiment, assessing the cost by the component of fluid modeling and the number of phase and the implicit level that is used for mathematics discretization are measured in simulator 200 operating process.Such as, as shown in Figures 4 and 5, exemplary reservoir model 400 has corresponding matrix equality 500, and wherein for IMPES node, the number of lines associated with each node is 1, equals the number of the component in CI district.IMPES refers to that hidden pressure shows saturated (implicit pressure explicit saturation), and CI refers to the implicit of coupling.In matrix equality 500, each nonzero element can be relevant with the floating-point operation being converted into some number assessed the cost.In special domain, more nonzero elements mean in special domain and more work.

In the exemplary embodiment, in model 100 and simulator 200, the method for parallelization provides timely and/or by calculating the adaptive unstructured grid 106 of classification so that the parallel performance of Optimized Simulated device.In the exemplary embodiment, parallel method can walk abreast or perform continuously.In the exemplary embodiment, parallel method can with the memory parallel machine shared, as multi-CPU available now/multinuclear desktop machine performs in simulator 200, because if data can local access and without network send or receive, then data map more effective again, but it can use on multiple available parallel machine, and these machines comprise, such as distributed memory bunch, memory chip (cell chip) and other multi core chips.

In the exemplary embodiment, when the parallel method data comprised in Confirming model 100 need again the tolerance of subregion, and it is different that this time calculates classification to the difference of the operating procedure of simulator 200.In the exemplary embodiment, parallel method comprises the different choice of the subregion of data in execution model 100.In the exemplary embodiment, according to the calculating that will perform in the operating procedure of simulator 200, the method for parallelization supplies a model the different subregions of data in 100.And, in the exemplary embodiment, for calculating classification given in operating procedure one or more in simulator 200, the model 100 that data partition best-fit dissimilar in corresponding model 100 is different.

In the exemplary embodiment, one or more parallel method comprise the following steps in one or more: 1) be the method for the data in Optimal Parallel solver algorithmic statement partition model 100; 2) based on the measurement of the inequality of balancing the load, the solver in the one or more operating procedures of zoning simulated device 200 and non-solver compute classes method for distinguishing; 3) based under list the subregion of data in dynamically adapting model: a) as the tolerance that an operation part for simulator calculates, as measured the iterations in flash calculation (flash calculation), then be transition front (phase-transition front), etc., ..., and/or performance running time of b) history and predictability; 4) provide correct node and connection weight to existing figure partition scheme; And/or 5) by multiple theory and/or heuristic, minimize facility and high yield area cutting.

With reference to figure 6a, in the exemplary embodiment, the operation of model 100 is simulated with reservoir simulator 600, performs well management 602 in simulator 600.In the exemplary embodiment, in model 100, the well management 602 of well 102 comprises iterative process 604, and wherein performing Jacobi's structure and flow rate calculation 606, is then linear solution 608 and property calculation 610.In the exemplary embodiment, after complete process 604, generate result/checkpoint I/O 612.

In the exemplary embodiment, as shown in Figure 6 b, whether the data of well management 602 in 602a in Confirming model 100 should rezone to improve treatment effeciency and/or well management precision.If the well management in model 100 should be made to assess the cost balancing the load again, then the data in 602b in model are rezoned, and the workload associated with well management in 602c can distribute between multiple CPU or CPU core.

In the exemplary embodiment, as fig. 6 c, whether the data in 606a in Jacobi's structure and flow rate calculation 606 Confirming model 100 should rezone to improve treatment effeciency and/or Jacobi constructs and the precision of flow rate calculation.If the data in model 100 should rezone, then the data in 606b in model are rezoned, and in 606c, construct the workload associated with flow rate calculation with Jacobi can distribute among multiple CPU or CPU core.

In the exemplary embodiment, as shown in fig 6d, whether the data solved in 608 Confirming model 100 at 608a neutral line should rezone to improve the precision for the treatment of effeciency and/or linear solution.If the data in model 100 should rezone, then in 608b, the data in model are rezoned, and the workload associated with linear solution in 608d can distribute among multiple CPU or CPU core.

In the exemplary embodiment, as shown in fig 6e, whether the data in 610a in property calculation 610 Confirming model 100 should rezone to improve the precision for the treatment of effeciency and/or property calculation.If the data in model 100 should rezone, then in 610b, the data in model are rezoned, and can distribute among multiple CPU or CPU core with the workload of linear properties compute associations in 610c.

In the exemplary embodiment, reservoir simulator 600 can be used such as, and one or more all-purpose computer, special-purpose computer, analog processor, digital processing unit, CPU and/or distributed computing system realize.

With reference to figure 7, above with reference to the one or more operating procedure 602b of Fig. 6 a to 6e explanation, the method 700 of 606b, 608b, 610b and/or 612b realization partition data in reservoir model 100, the data wherein in this model are prepared and/or revise to be input in subregion process 702.In the exemplary embodiment, the data of preparation/correction model 100 so as 702 be input to subregion process comprise following in one or more: determine/revise node and connection weight from model and/or model streamline trace (tracing), or be any other partitioning algorithm preparation/Correction and Control parameter.In 704, model 100 prepare data be then by by the node of model in separate domains be connected subregion and be partitioned 704.In the exemplary embodiment, in 704 by the node of model 100 be connected subregion and comprise the figure subregion of model.After the subregion completing model 100, level and smooth and projection perform subregion in 706 after.Then in 708 by the subregion quality of one or more quality metric Confirming model 100.If the quality metric instruction of the subregion of model 100 calculates the data partition of poor efficiency, then method repeats step 702 to 710, until meet the subregion quality metric of model 100 in 710.

In the exemplary embodiment, as shown in Figure 8, the method 800 of model 100 subregion is included in the figure of coloring models in 802 to find isolated or close isolated node group.Such as, as shown in Figure 8 a, reservoir model 802a comprises multiple node 802b, and node 802b defines one or more pieces of 802c of associated nodes.In the exemplary embodiment, in 802, each piece of 802c of model 802a is colored, to find isolated or close isolated node group.In the exemplary embodiment, the instruction of the computational activity level required by the painted operation being simulation model 100 in specific piece of block 802c in 802.In the exemplary embodiment, in 802, the painted color rendering intent of block 802c provides, certain other equivalence or similar tolerance of some color reaction transfer rate (transmissibility) or conductivity (conductivity) in this color rendering intent, as Jacobi's pressure equation off-diagonal, namely model 100 isolated or the instruction close to isolated region.

In the exemplary embodiment, method 800 then in 804 by size chosen node color lump.Such as, as shown in Figure 8 b, color lump 802c be in 804 from left to right by size select and be designated block 802c1,802c2,802c3,802c4,802c5,802c6 and 802c7 respectively.

In the exemplary embodiment, the then node of each painted block with selecting of method 800 weighting in 806, thus illustrate that associate from each node of process in the simulation process of model 100 different assess the cost.Such as, as shown in Figure 8 c, in 806, the node of painted block 802c is according to the processing cost weighting of association, this processing cost with in the simulation process of model 100, process each node associate.

In the exemplary embodiment, then method 800 distributes the node of weighting to territory in 808, so that Optimization Work balancing the load.Such as, as shown in figure 8d, in 808, the method allocation block 802c1 and 802c7 is to territory 0, and allocation block 802c2,802c3,802c4,802c5 and 802c6 are to territory 1.

As shown in Figure 9, in the exemplary embodiment, by the method 900 of model 100 subregion, the calculated level be included in 902 required by each node selects the node in model on Descartes direction.Such as, as illustrated in fig. 9, reservoir model 902a comprises the well 902b, the region 902c that assess the cost higher and the region 902d assessing the cost lower.As shown in figure 9b, in 902, the node 902e in model 902a selects on given Descartes direction according to the calculated level associated with each node.

In the exemplary embodiment, then the calculating weight factor of all node 902e is added by method 900, thus for model 902a determines that the accumulation of grid calculates weight in 904.

In the exemplary embodiment, then method 900 in 906 distribution node 902e to special domain, until the accumulation of this special domain calculates the accumulation that weight equals grid calculate the predetermined percentage of weight.Such as, as is shown in fig. 9 c, in the exemplary embodiment, in 906, the node 902e in model 902a is assigned to territory 906a and 906b.

In the exemplary embodiment, as shown in figure 9d, if selecting of method 900 interior joint 902e performs in the X direction, so final domain 906a and 906b is generated.In the exemplary embodiment, alternatively, as shown in figure 9e, if in method 900, selecting of node 902e performs in the Y direction, then generate final domain 906a and 906b.

In the exemplary embodiment, method 900 then implementation quality inspection thus to determine in 906 whether selected subregion is suitable according to predetermined quality control standard in 908 and 910.

In several exemplary embodiment, available any direction of selecting of method 900 interior joint 902e provides, as x, y or z.And in the exemplary embodiment, the subregion in selected direction and selected territory can be iterative process, being uniformly distributed of the process of this iterative process Optimized model 902a.

In the exemplary embodiment, as shown in Figure 10, the method 1000 of model 100 subregion is included in the velocity field of Confirming model in 1002.Then the method 1000 velocity field tracking velocity streamline that basis is determined for model 100 in 1004.Such as, as shown in Figure 10 a, reservoir model 1004a comprises well 1004b and 1004c, and the streamline 1004d extended between well.

In the exemplary embodiment, then method 1000 projects up and down in vertical direction streamline thus generate stream curtain or streamline curtain in 1006.Such as, as shown in fig. lob, in 1006, streamline 1004d upwards projected and projects downwards thus generate stream curtain 1006a.

In the exemplary embodiment, then method 1000 extends stream curtain to the border of model meshes in 1008, and adjustment simultaneously flows curtain with away from well.Such as, as shown in figure l oc, in 1008, adjustment is flowed curtain 1006a thus is generated stream curtain 1008a, and it extends to the border of the grid of model 1004a, simultaneously away from well 1004b and 1004c.As a result, model 1004a is divided into territory 1008b and 1008c.

In the exemplary embodiment, optimally partitioned then with multiple stream curtain preference pattern 100 in 1010 of method 1000.

In the exemplary embodiment, method 1000 then in 1012 implementation quality inspection thus according to predetermined quality control standard determine in 1010 select subregion whether be suitable.According to predetermined quality control standard, if the subregion selected in 1010 is inappropriate, then method continues iterative modifications subregion until it is suitable.

In the exemplary embodiment, model 100 subregion can be minimized the processing cost with simulator 600 simulation model by using method 1000.Especially, in the exemplary embodiment, because velocity profile can the dynamic flow of fluid in approximate model 100, therefore streamline represents and can to jump the border of the impact caused by minimum material attribute thereon.

In the exemplary embodiment, as shown in figure 11, the method 1100 of model 100 subregion is determined node and the connection weight factor in 1102, or is if necessary revised.In the exemplary embodiment, the node weights factor represents the processing cost associated with model 100 interior joint, and the connection weight factor represents the degree which node is connected to other nodes.

In the exemplary embodiment, method 1100 then in 1106 according to the node weights determined in 1102 and the node connection weight factor by model 100 subregion.In the exemplary embodiment, in 1106, model 100 is partitioned thus in multiple territory, is uniformly distributed the processing cost of simulation model.In the exemplary embodiment, in 1106, the connection between the node of strong ties in Slicing Model for Foreign is avoided in the territory of the model 100 of structure.

In the exemplary embodiment, method 1100 then implementation quality inspection thus to determine in 1106 whether selected subregion is suitable according to predetermined quality control criterion in 1108.

In the exemplary embodiment, be variable in time in the determination of 1102 interior joint weight factors and/or the connection weight factor.

In the exemplary embodiment, in 1102, determine that the connection weight factor is by determining that node and the distance recently between well perform.Such as, as shown in figure 12, model 100 interior joint can indicate the respective distance of they and its well 102 recently separately with color coding.Node 104 can be used as with the distance of nearest well the part determining connection weight in 1102.In the exemplary embodiment, node 104 from well 102 more close to, connection weight is higher, therefore in 1106 of method 1100 by the process of model 100 subregion, disconnect this node and the connection recently between well is more unworthy.

In the exemplary embodiment, as shown in figure 13, the method 1300 of model 100 subregion is determined in 1301 or revise partitioned parameters and in 1302 the subregion of generation model 100.In the exemplary embodiment, then method 1300 determines the distance from the border of the subregion generated to adjacent well 102 model 100 in 1304.If be less than a certain predetermined value with the distance on the border of any one subregion generated in 1306, then repeat 1301 to 1306 until be more than or equal to this predetermined value with the distance on all borders of the subregion generated.

In the exemplary embodiment, as shown in figure 14, the method 1400 of model 100 subregion is determined partitioned parameters in 1401, if or need, revise these parameters, and generate subregion in 1402.In the exemplary embodiment, as shown in figures 14a, in 1402, model 1402a subregion is territory 1402b, 1402c and 1402d by method 1400.

In the exemplary embodiment, method 1400 all nodes of then determining along partition model between territory border in 1404.In the exemplary embodiment, as shown in fig. 14b, in 1404, method 1400 determines the node 1402bc fallen along the border between 1402b and 1402c of territory, the node 1402cd fallen along the border between 1402c and 1402d of territory, the node 1402db fallen along the border between 1402d and 1402b of territory.

In the exemplary embodiment, method 1400 then in 1406 projected boundary node to plane and matching by the curve of boundary node of projection.In the exemplary embodiment, as shown in figure 14 c, method 1400 in 1406 projected boundary node 1402bc, 1402cd and 1402bd to X-Y plane and matching respectively by curve 1406a, 1406b and 1406c of boundary node of projection.

In the exemplary embodiment, method 1400 then in 1408 from generate curve extend another direction projection smooth surface, this direction can be orthogonal with plane selected in 1408, as shown in Figure 14 d, in the exemplary embodiment, the method projects smoothed curve 1408a, 1408b and 1408c in z-direction.As a result, model 1402 is partitioned into territory 1408aa, 1408bb and 1408cc.

In the exemplary embodiment, method 1400 then in 1410 Confirming model 1402a whether divide the quality of separate domains into enough.

With reference to Figure 15, the exemplary embodiment of the method 1500 of reservoir model subregion is generated in 1502 the subregion of reservoir model.In the exemplary embodiment, then method 1500 compares the calculated performance of the calculated performance generating subregion of reservoir model and the historical data about reservoir model subregion in 1504.In the exemplary embodiment, method 1500 then in 1506 iteration use the difference of the calculated performance of the calculated performance generating subregion of reservoir model and the historical data of reservoir model subregion, thus improve the subregion of reservoir model.In the exemplary embodiment, then method 1500 determines to be whether the quality of separate domains is enough by reservoir model subregion in 1508.If poor quality, then attempt new partition method.

In the exemplary embodiment, method 1500 uses the quality of the subregion of the one or more staticametric determination reservoir models of subregion quality in 1508, and these staticametric such as can comprise the smoothness metric on the statistical measures of territory contour connection, the average and standard deviation of transfer rate, Jacobi's off-diagonal element, subregion internal area border.In the exemplary embodiment, the smoothness metric on border, territory such as can pass through the boundary node of specific interface between projection adjacent domains in plane, and matching is provided by the curve of this projection.In the exemplary embodiment, the degree of curve projection provides the instruction of the heterogeneous degree in border between adjacent domains.

In the exemplary embodiment, in method 700, by the grid node of model 100 and the one or more aspects being connected subregion and comprising to territory partition method 800 and/or 900 and/or 1000 and/or 1100 and/or 1300 and/or 1400 and/or 1500.

In the exemplary embodiment, one or more in the operation of simulator 600 and/or method 600,700,800,900,1000,1100,1300,1400 and/or 1500 is implemented thus the treatment effeciency of simulation with the performance metric Optimized Simulated reservoir 100 below one or more further: 1) solver iteration convergence rate; 2) real time (wall clock time) spent and the ratio of CPU; 3) property calculation; And/or 4) Jacobi structure and flow rate calculation.

In the exemplary embodiment, the total degree of the outer iteration of linear iteraction solver is the good indicator of parallel efficiency and partitioning problem.

In the exemplary embodiment, in continuous (serial) processing procedure of simulator 600, CPU should equal elapsed time amount for the time quantum calculated, that is, the real time of cost.In the exemplary embodiment, in the parallel process of simulator 600, all CPU work to total work for the treatment of that simulation performs should keep ideally with run continuously the same, but the real time of the cost of disappearance should reduce and the real time spent to the ratio of CPU number---the real time of cost and the ratio of CPU should be proportional with 1/ (CPU number).But if CPU usage increases soon than the real time of cost, then the ratio of the real time spent to CPU also should decline.Such as, if parallel processing is more much lower than the efficiency of serial process, then this may occur.If used in conjunction with other tolerance, then the ratio of the real time spent to the CPU time is the useful dynamic measurement of parallel efficiency.Such as, along with Simulation Progress, the change of real time to the ratio of CPU of cost is the instruction of problem.In the exemplary embodiment, similar reservoir simulation model can be expected and run similarly.Especially, for similar physical reservoir model, the parallel performance of expection reservoir model simulation is similar.Therefore, we can more current cost real time to the CPU time than and real time of current cost of similar reservoir model to CPU time ratio, and infer parallel efficiency.

In the exemplary embodiment, the one or more methods being independent territory by reservoir model 100 subregion illustrated above with reference to Fig. 1-15 perform, in a dynamic fashion at least to minimize the degeneration with the reservoir simulator parallel processing efficiency of time correlation.

In the exemplary embodiment, the balancing the load of the simulation of reservoir model is inferred by other tolerance of CPU workload.Particularly, the different classes of calculating performed in reservoir model simulation process has different useful metrics for assessing the cost of every grid node.

In the exemplary embodiment, state equation (EOS) property calculation once calculates a node usually in the process of simulation reservoir model, produces the work tolerance performed in flash calculation process.Flash calculation is the process based on input pressure and component determination fluid volume and composition.This tolerance can adopt the form of flash distillation solver iteration count---to distinguish with the linear matrix equation solver of whole system.In the exemplary embodiment, another tolerance of flash calculation cost is fluid complexity---mutually how many and fluid components is being present in node instantaneously.This has additional benefit, namely all has applicability for EOS and dirty oil (BO) fluid model.

In the exemplary embodiment, Jacobi's structure and fluid calculation, be generally the cost of vector-vector matrix-vector operation, and the number by component and phase and the implicit level tolerance for mathematics discretization, fluid is by these components and phase modeling.In the exemplary embodiment, for by the phase of fluid modeling and component more, then the state variable that must calculate is more.In the exemplary embodiment, higher at the implicit degree of given node to model attributes, then the cost of node calculate is higher.The implicity of calculation requirement is higher, then derivative calculations is more.

In the exemplary embodiment, as shown in figure 16, for each CPU associated with linear solution, the operating efficiency of the parallel processing of reservoir simulator linear solution is by checking that the solver time 1600 is assessed.

In the exemplary embodiment, the framework using partitioning algorithm to optimize reservoir simulator parallel performance comprises: a) be given partitioning algorithm adjustment parameter---as being figure partitioning algorithm (GPA) computing node and the connection weight factor; B) partition method selected is run, such as GPA; C) post processing improvement-repairing (fix-up) and level and smooth subregion is carried out; D) subregion quality is assessed; If with e) quality can accept, then exit, otherwise the parameter suitably changing GPA carrys out repetitive process.

In the exemplary embodiment, the every class performed in the operating process of simulator 600 calculates can be benefited from the target partition method of himself, and devises the partition scheme being exclusively used in every class and calculating.

In the exemplary embodiment, because the physics of reservoir simulation and mathematical properties are normally with time correlation, so partition method described here is independence self-adapting---that is, the partition scheme that the every class performed during the operation of simulator 600 calculates can adjust with himself target frequency.

In the exemplary embodiment, partition method described here adopts the tolerance of physical base to determine subregion quality.

In the exemplary embodiment, model 100 subregion is comprised geometry Slicing Model for Foreign; The painted model node selected---based on the physics weight of node and connection, wherein the connection of mid-span has physical base threshold value; With flow base subregion.In the exemplary embodiment, flow base subregion comprises based on streamline, such as streamline curtain, stream pipe cohesion (stream tube agglomeration) and level and smooth; With with based on flow or coefficient weight zoning plan thus minimize the large jump of coefficient on cross-domain border.

Described the method for simulation reservoir model, the method comprises generation reservoir model; Be multiple territory by the reservoir model subregion of generation; And simulate the reservoir model of subregion.In the exemplary embodiment, simulation reservoir model comprises division reservoir simulation is multiple processing element; Processing element multiple with parallel processing.In the exemplary embodiment, the multiple processing element of parallel processing comprises and rezones the reservoir model of generation into multiple territory.In the exemplary embodiment, rezoned by the reservoir model of generation as multiple territory comprises: a) pretreatment reservoir model, it can include but not limited to select/change partitioning algorithm, and for select partitioning algorithm to determine/corrected parameter; B) be multiple territory by the reservoir model subregion of generation; C) post processing subregion reservoir model thus correct the reservoir model of subregion; D) quality of the subregion reservoir model of post processing is assessed; And if the quality of e) the subregion reservoir model of post processing is less than predetermined value, then repeat a, b, c, d and e.In the exemplary embodiment, simulate reservoir model to comprise reservoir model is rezoned; The simulation of reservoir is divided into multiple processing element and the multiple processing element of parallel processing.In the exemplary embodiment, be that multiple territory comprises a) pretreatment reservoir model by the reservoir model subregion of generation, this can include but not limited to select/change partitioning algorithm and for select partitioning algorithm to determine/corrected parameter; B) be multiple territory by the reservoir model subregion of generation; C) post processing subregion reservoir model thus correct the reservoir model of subregion; D) quality of the subregion reservoir model of post processing is assessed; And if the quality of e) the subregion reservoir model of post processing is lower than predetermined value, then repeat a, b, c, d and e.In the exemplary embodiment, be that multiple territory comprises the reservoir model of painted generation thus generates the joint block with corresponding colour coding (color code), the degree that this colour coding representation node block and other joint blocks are isolated by the reservoir model subregion of generation; Select the joint block of this color dyes; The color-coded joint block that weighting is selected thus the processing cost that associates with each node is described; Corresponding territory is given with the joint block of weights assigned.In the exemplary embodiment, by the reservoir model subregion of generation be the processing cost level that associates of node that multiple territory comprises in the reservoir model determined and generate; According to the processing cost associated with node chosen node in one direction; The processing cost of the node selected in the direction in which is added thus determines the total processing cost with this directional correlation; Give corresponding territory with the node distributed in the direction in which thus distribute total processing cost in the direction in which.In the exemplary embodiment, be that multiple territory comprises the velocity field that associates of reservoir model determined and generate by the reservoir model subregion of generation; Follow the tracks of the streamline associated with velocity field; Projection streamline thus generate stream curtain; And extend stream curtain to generation reservoir model border thus be territory by the reservoir model subregion of generation.In the exemplary embodiment, be wherein that multiple territory comprises the border of extension stream curtain to the reservoir model generated further by the reservoir model subregion of generation, thus be territory by the reservoir model subregion of generation, avoid border crossing with the well defined in the reservoir model of generation simultaneously.In the exemplary embodiment, be that multiple territory comprises the multiple stream curtain of generation further and extends the border of stream curtain to the reservoir model generated by the reservoir model subregion of generation, thus be that multiple territory is gathered by the reservoir model subregion of generation.In the exemplary embodiment, by the reservoir model subregion of generation be multiple territory comprise further determine to gather with multiple territory in the processing cost of each territory set associative distribute; From multiple territories set with the distribution of optimization process cost, subregion is selected with the reservoir model for generating.In the exemplary embodiment, by the reservoir model subregion of generation be the processing cost that multiple territory comprises each node of reservoir model of determining and generate and associates; Determine generate reservoir model each node between connection level; Be multiple territory according to the processing cost determined with being communicated with the reservoir model subregion of level by generation.In the exemplary embodiment, be that multiple territory is included in the processing cost being uniformly distributed between territory and determining according to the processing cost determined with being communicated with the reservoir model subregion of level by generation.In the exemplary embodiment, according to the processing cost determined be communicated with the reservoir model subregion of level by generation be multiple territory comprise grouping same territory in the node of connection level more than predeterminated level.In the exemplary embodiment, by the reservoir model subregion of generation be the distance that multiple territory comprises between the adjacent well determining to define in the reservoir model on border, territory and generation; As requested, according to the distance determined rezone generate reservoir model.In the exemplary embodiment, be that multiple territory comprises all nodes along boundary alignment between territory in the reservoir model determining to generate by the reservoir model subregion of generation; Projected boundary node is to plane and the curve of boundary node of matching by having projected; And and the territory of reservoir model that generates of the orthogonal direction Projective Curve of matched curve thus definition between border.In the exemplary embodiment, be that multiple territory comprises the subregion of the reservoir model comparing generation and previous reservoir model subregion by the reservoir model subregion of generation.

Described the method for simulation reservoir model, the method comprises generation reservoir model; Be multiple territory by the reservoir model subregion of generation; What divide reservoir is modeled as multiple processing element; The multiple processing element of parallel processing; And be other multiple territory by the reservoir model subregion of generation at least one times in parallel process; Be wherein that multiple territory comprises by the reservoir model subregion of generation: a) pretreatment reservoir model, it can include but not limited to select/change partitioning algorithm, and for select partitioning algorithm to determine/corrected parameter; B) be multiple territory by the reservoir model subregion of generation; C) post processing subregion reservoir model thus correct the reservoir model of subregion; D) quality of the subregion reservoir model of post processing is assessed; And if the quality of e) the subregion reservoir model of post processing is less than predetermined value, then repeat a, b, c, d and e.

Described the method for simulation reservoir model, the method comprises generation reservoir model; Be multiple territory by the reservoir model subregion of generation; What divide reservoir is modeled as multiple processing element; The multiple processing element of parallel processing; And be other multiple territory by the reservoir model subregion of generation at least one times in parallel process; Be wherein that multiple territory comprises the level determining the processing cost associated with the reservoir model interior nodes generated by the reservoir model subregion of generation; According to the processing cost chosen node associated with node; The processing cost of the node selected is added thus determines total processing cost of associating with node; With distribution node is given corresponding territory thus distribute total processing cost between territory.

Described and be included in for simulating the computer program of reservoir model in tangible medium, it comprise instruction for: generate reservoir model; Be multiple territory by the reservoir model subregion of generation; With the reservoir model of simulation subregion.In the exemplary embodiment, simulate reservoir model comprise divide reservoir be modeled as multiple processing element and the multiple processing element of parallel processing.In the exemplary embodiment, the multiple processing element of parallel processing comprises and rezones the reservoir model of generation into multiple territory.In the exemplary embodiment, rezoned by the reservoir model of generation as multiple territory comprises a) pretreatment reservoir model, it can include but not limited to select/change partitioning algorithm, and for select partitioning algorithm to determine/corrected parameter; B) be multiple territory by the reservoir model subregion of generation; C) post processing subregion reservoir model thus correct the reservoir model of subregion; D) quality of the subregion reservoir model of post processing is assessed; And if the quality of e) the subregion reservoir model of post processing is less than predetermined value, then repeat a, b, c, d and e.In the exemplary embodiment, simulate reservoir model to comprise reservoir model is rezoned; What divide reservoir is modeled as multiple processing element; Processing element multiple with parallel processing.In the exemplary embodiment, be that multiple territory comprises a) pretreatment reservoir model by the reservoir model subregion of generation, this can include but not limited to select/change partitioning algorithm and for select partitioning algorithm to determine/corrected parameter; B) be multiple territory by the reservoir model subregion of generation; C) post processing subregion reservoir model thus correct subregion; D) quality of the subregion reservoir model of post processing is assessed; And if e) the quality of the reservoir model of post processing subregion lower than predetermined value, then repeat a, b, c, d and e.In the exemplary embodiment, be that multiple territory comprises the reservoir model of painted generation thus generates the joint block with corresponding colour coding, the degree that this colour coding representation node block and other joint blocks are isolated by the reservoir model subregion of generation; Select the joint block of this color dyes; The color-coded joint block that weighting is selected thus the processing cost that associates with each node is described; Corresponding territory is given with the joint block of weights assigned.In the exemplary embodiment, by the reservoir model subregion of generation be the processing cost level that associates of node that multiple territory comprises in the reservoir model determined and generate; According to the processing cost associated with node chosen node in one direction; The processing cost of the node selected in the direction in which is added thus determines the total processing cost with this directional correlation; Give corresponding territory with the node distributed in the direction in which thus distribute the party's total processing cost upwards.In the exemplary embodiment, be that multiple territory comprises the velocity field that associates of reservoir model determined and generate by the reservoir model subregion of generation; Follow the tracks of the streamline associated with velocity field; Projection streamline thus generate stream curtain; And extend stream curtain to generation reservoir model border thus be territory by the reservoir model subregion of generation.In the exemplary embodiment, be that multiple territory comprises the border of extension stream curtain to the reservoir model generated further by the reservoir model subregion of generation, thus be territory by the reservoir model subregion of generation, avoid border crossing with the well defined in the reservoir model generated simultaneously.In the exemplary embodiment, be that multiple territory comprises the multiple stream curtain of generation further by the reservoir model subregion of generation; Flow the border of curtain to the reservoir model generated with extending, thus be the set of multiple territory by the reservoir model subregion of generation.In the exemplary embodiment, by the reservoir model subregion of generation be multiple territory comprise further determine to gather with multiple territory in the processing cost of each territory set associative distribute; From multiple territories set with the distribution of optimization process cost, subregion is selected with the reservoir model for generating.In the exemplary embodiment, the reservoir model subregion of generation is comprised the processing cost that each node of reservoir model of determining and generate associates to multiple territory; Determine generate reservoir model each node between connection level; Be multiple territory according to the processing cost determined with being communicated with the reservoir model subregion of level by generation.In the exemplary embodiment, be that multiple territory is included in the processing cost being uniformly distributed between territory and determining according to the processing cost determined with being communicated with the reservoir model subregion of level by generation.In the exemplary embodiment, be that multiple territory comprises connective node more than predeterminated level in the same territory of grouping according to the processing cost determined with being communicated with the reservoir model subregion of level by generation.In the exemplary embodiment, by the reservoir model subregion of generation be the distance that multiple territory comprises between the adjacent well determining to define in the reservoir model on border, territory and generation; As requested, according to the distance determined rezone generate reservoir model.In the exemplary embodiment, by the reservoir model subregion of generation be all nodes that multiple territory comprises along boundary alignment between territory in the reservoir model determining to generate; Projected boundary node to plane and matching by the curve of boundary node of projection; And and the territory of reservoir model that generates of the orthogonal direction Projective Curve of matched curve thus definition between border.In the exemplary embodiment, be that multiple territory comprises the subregion of the reservoir model comparing generation and previous reservoir model subregion by the reservoir model subregion of generation.

Described and be included in for simulating the computer program of reservoir model in tangible medium, it comprise instruction for: generate reservoir model; Be multiple territory by the reservoir model subregion of generation; What divide reservoir is modeled as multiple processing element; The multiple processing element of parallel processing; In parallel process, be other multiple territory by the reservoir model subregion of generation at least one times; Be wherein that multiple territory comprises by the reservoir model subregion of generation: a) pretreatment reservoir model, it can include but not limited to select/change partitioning algorithm, and for select partitioning algorithm to determine/corrected parameter; B) be multiple territory by the reservoir model subregion of generation; C) post processing subregion reservoir model thus correct the reservoir model of subregion; D) quality of the subregion reservoir model of post processing is assessed; And if the quality of e) the subregion reservoir model of post processing is less than predetermined value, then repeat a, b, c, d and e.

Described and be included in for simulating the computer program of reservoir model in tangible medium, it comprise instruction for: generate reservoir model; Be multiple territory by the reservoir model subregion of generation; What divide reservoir is modeled as multiple processing element; The multiple processing element of parallel processing; And in parallel process, at least one times by generate reservoir model subregion be other multiple territories; It is wherein that multiple territory comprises by the reservoir model subregion of generation: the level determining the processing cost associated with the reservoir model interior nodes generated; According to the processing cost chosen node associated with node; The processing cost of the node selected is added thus determines total processing cost of associating with node; And distribution node is given corresponding territory thus distribute total processing cost between territory.

Described the system of simulation reservoir model, it comprises the device generating reservoir model; By the device that the reservoir model subregion of generation is multiple territory; The device of the reservoir model of simulation subregion.In the exemplary embodiment, the device of simulating reservoir model comprises the device simulation of reservoir being divided into multiple processing element; With the device of the multiple processing element of parallel processing.In the exemplary embodiment, the device of the multiple processing element of parallel processing comprises and rezones the reservoir model of generation into the device in multiple territory.In the exemplary embodiment, the device rezoned by the reservoir model of generation as multiple territory comprises the device of a) pretreatment reservoir model, and it can include but not limited to the parameter selected/change partitioning algorithm and determine/revise to have selected partitioning algorithm; B) by the reservoir model subregion of generation be the device in multiple territory; C) post processing subregion reservoir model thus correct the device of reservoir model of subregion; D) device of the quality of the subregion reservoir model of post processing is assessed; And e) determine whether the quality of subregion reservoir model of post processing is less than the device of predetermined value, if it is repeats a, b, c, d, and the device of e.In the exemplary embodiment, the device of simulating reservoir model comprises the device rezoned by reservoir model; The simulation of reservoir is divided into the device of the device of multiple processing element and the multiple processing element of parallel processing.In the exemplary embodiment, the device being multiple territory by the reservoir model subregion of generation comprises the device of a) pretreatment reservoir model, and it can include but not limited to the parameter selected/change partitioning algorithm and determine/revise to have selected partitioning algorithm; B) by the reservoir model subregion of generation be the device in multiple territory; C) post processing subregion reservoir model thus correct the device of reservoir model of subregion; D) device of the quality of the subregion reservoir model of post processing is assessed; And e) determine whether the quality of subregion reservoir model of post processing is less than the device of predetermined value, if it is repeats a, the device of b, c, d and e.In the exemplary embodiment, the device being multiple territory by the reservoir model subregion of generation comprises the reservoir model of painted generation thus generates the device with the joint block of corresponding colour coding, the degree that this colour coding representation node block and other joint blocks are isolated; Select the device of the joint block of color dyes; The color-coded joint block that weighting is selected thus the device of processing cost associated with each node is described; With the joint block of weights assigned to the device of corresponding domain.In the exemplary embodiment, the device being multiple territory by the reservoir model subregion of generation comprises, and determines the device of the processing cost level associated with the reservoir model interior nodes generated; According to the device of the processing cost associated with node chosen node in one direction; The processing cost of the node selected in the direction in which is added thus determines the device with total processing cost of this directional correlation; Distribute the party's node upwards to corresponding domain and distribute the device of the party's upwards total processing cost.In the exemplary embodiment, the device being multiple territory by the reservoir model subregion of generation comprises the device determining the velocity field associated with the reservoir model generated; Follow the tracks of the device of the streamline associated with velocity field; Projection streamline thus generate stream curtain device; With extend stream curtain to generation reservoir model border thus be the device in territory by the reservoir model subregion of generation.In the exemplary embodiment, the device being multiple territory by the reservoir model subregion of generation comprise further extend stream curtain to generation reservoir model border thus be the device that territory avoids border crossing with the well defined in the reservoir model of generation simultaneously by the reservoir model subregion of generation.In the exemplary embodiment, the device being multiple territory by the reservoir model subregion of generation comprises the device generating multiple stream curtain further; Flowing curtain to the reservoir model border generated thus by the reservoir model subregion of generation with extension is the device that multiple territory is gathered.In the exemplary embodiment, the device being multiple territory by the reservoir model subregion of generation comprise further determine to gather with multiple territory in the device that distributes of the processing cost of each territory set associative; With the device from multiple territories set with the distribution of optimization process cost being the reservoir model selection subregion generated.In the exemplary embodiment, the device being multiple territory by the reservoir model subregion of generation comprises the device determining the processing cost associated with each node of the reservoir model generated; The device of connection level between each node determining the reservoir model generated; With according to the processing cost determined be communicated with the device that the reservoir model subregion of level by generation is multiple territory.In the exemplary embodiment, be that the device in multiple territory is included in the device being uniformly distributed the processing cost determined between territory according to the processing cost determined with being communicated with the reservoir model subregion of level by generation.In the exemplary embodiment, be the device that the device in multiple territory comprises the node of connection level more than predeterminated level in grouping same domain according to the processing cost determined with being communicated with the reservoir model subregion of level by generation.In the exemplary embodiment, the device being multiple territory by the reservoir model subregion of generation comprises the device of the spacing determining the adjacent well defined in the reservoir model on border, territory and generation; To rezone the device of reservoir model generated according to the distance determined on request.In the exemplary embodiment, the device being multiple territory by the reservoir model subregion of generation comprises the device along all nodes of boundary alignment between territory in the reservoir model determining to generate; Projected boundary node passes through the device of the curve of the boundary node of projection to plane and matching; And the device on border in the direction Projective Curve orthogonal with matched curve thus between the territory of reservoir model that generates of definition.In the exemplary embodiment, the device being multiple territory by the reservoir model subregion of generation comprises the subregion of the reservoir model comparing generation and the device of previous reservoir model subregion.

The system of the simulation reservoir model illustrated comprises the device generating reservoir model; By the device that the reservoir model subregion of generation is multiple territory; Divide the device being modeled as multiple processing element of reservoir; The device of the multiple processing element of parallel processing; With in parallel process at least one times by device that the reservoir model subregion of generation is other multiple territory; The device being wherein multiple territory by the reservoir model subregion of generation comprises: a) device of pretreatment reservoir model, and it can include but not limited to select/change partitioning algorithm, and for select partitioning algorithm to determine/corrected parameter; B) by the reservoir model subregion of generation be the device in multiple territory; C) post processing subregion reservoir model thus correct the device of reservoir model of subregion; D) device of the quality of the subregion reservoir model of post processing is assessed; And e) determine whether the quality of subregion reservoir model of post processing is less than the device of predetermined value, if it is repeats a, the device of b, c, d and e.

The system of the simulation reservoir model illustrated comprises the device generating reservoir model; By the device that the reservoir model subregion of generation is multiple territory; Divide the device being modeled as multiple processing element of reservoir; The device of the multiple processing element of parallel processing; With in parallel process at least one times by device that the reservoir model subregion of generation is other multiple territory; The device being wherein multiple territory by the reservoir model subregion of generation comprises: the device determining the processing cost level associated with the reservoir model interior nodes generated; According to the device of the processing cost chosen node associated with node; The processing cost of the node selected is added thus determines the device of the total processing cost associated with node; With distribution node distributes the device of total processing cost to corresponding domain between territory.

Be appreciated that can not depart from scope makes change to the present invention.Such as, the instruction of illustrative embodiment of the present invention can be used to improve the computational efficiency of the n dimension computer model of the other types comprising network.

Although show and described illustrative embodiment of the present invention above, contemplate the amendment of wide region in foregoing disclosure content, change and replacement.In some cases, some feature of the present invention can adopt when not corresponding other features of use.Therefore, in a wide range and to understand claim in the mode consistent with scope be suitable.

Claims

1. simulate a method for reservoir model, it comprises:

Generate described reservoir model, described reservoir model is to the fluid, energy and/or the gas modeling that flow in hydrocarbon reservoir, well and relevant earth's surface facility;

By multiple set that generated reservoir model subregion is not same area, each territory corresponds to the subregion of the specific part of described reservoir model, the parallel processing efficiency of the simulation of reservoir model described in described partition zone optimizing;

Need, by balancing the load again, dynamically generated reservoir model to be rezoned into multiple territory during the simulation of described reservoir model, to improve parallel performance in response to determining to simulate assessing the cost of described reservoir model;

The operating procedure that the simulation being divided in described reservoir model comprises is multiple processing element; And

Rezone based on described, multiple processing element described in parallel processing;

Analog result according to described reservoir model controls production of hydrocarbons activity.

2. method according to claim 1, wherein rezones generated reservoir model as multiple territory comprises:

A) by selecting partition scheme and determining that the parameter of described partition scheme carrys out reservoir model described in pretreatment;

B) described partition scheme is used to be multiple territory by generated reservoir model subregion;

C) post processing subregion reservoir model thus improve further the parallel performance of the calculating of institute's subregion;

D) quality of the subregion reservoir model of post processing is assessed; And

If e) quality of the subregion reservoir model of post processing is less than predetermined value, then repeat a, b, c, d by the partition scheme revised and parameter, and e.

3. generated reservoir model subregion is wherein that multiple territory comprises by method according to claim 1:

Identify the block of the node be isolated from each other;

Select identified joint block by size;

The joint block weighting selected is illustrated the processing cost that associates with each piece;

The joint block of weighting is selected based on processing cost; And

The joint block of weights assigned gives corresponding territory.

4. generated reservoir model subregion is wherein that multiple territory comprises by method according to claim 1:

Determine the level of the processing cost associated with each node in the reservoir model generated;

According to the level of the processing cost determined, geometric direction is selected described node;

The calculating weight factor of the node selected is added;

And distribution node to territory until the accumulation weight in described territory equals the selected percentage of described calculating weight factor sum.

5. generated reservoir model subregion is wherein that multiple territory comprises by method according to claim 1:

Determine and the velocity field that generated reservoir model associates;

Follow the tracks of the streamline associated with described velocity field;

Project described streamline thus generate streamline curtain; And

Extending described streamline curtain to the border of generated reservoir model thus by generated reservoir model subregion is territory.

6. generated reservoir model subregion is wherein that multiple territory comprises by method according to claim 1:

Determine and the processing cost that each node of generated reservoir model associates;

Determine each internodal processing cost being communicated with horizontal relevance with generated reservoir model; And

According to determined processing cost, be multiple territory by generated reservoir model subregion.

7. method according to claim 6 is wherein that multiple territory comprises according to determined processing cost with being communicated with level by generated reservoir model subregion:

The node grouping of level more than predeterminated level will be communicated with in same domain.

8. generated reservoir model subregion is wherein that multiple territory comprises by method according to claim 1:

By described territory subregion;

Determine the distance between the border in described territory and the interior adjacent well defined of reservoir model of generation; And

As required, to rezone generated reservoir model according to determined distance, thus therefore mobile domains subregion improves described solver performance away from described well.

9. method according to claim 8, wherein comprises described territory subregion:

Identify the block of the node be isolated from each other;

Select identified joint block by size;

The joint block that weighting is selected thus consider the processing cost that associates with each piece;

The joint block of weighting is selected based on processing cost; And

Distribute the joint block of weighting to corresponding territory.

10. method according to claim 8, wherein comprises described territory subregion:

Determine the level of the processing cost associated with each node in generated reservoir model;

The calculating weight factor of the node selected is added;

11. methods according to claim 8, wherein comprise described territory subregion:

Determine and the velocity field that generated reservoir model associates;

Follow the tracks of the streamline associated with described velocity field;

Project described streamline thus generate streamline curtain; And

12. methods according to claim 8, wherein comprise described territory subregion:

Determine and the processing cost that each node in generated reservoir model associates;

Determine the processing cost being communicated with horizontal relevance between each node of generated reservoir model; And

Be multiple territory according to determined processing cost by generated reservoir model subregion.

Generated reservoir model subregion is wherein that multiple territory comprises by 13. methods according to claim 1:

By described territory subregion;

The all nodes in generated reservoir model are determined along the border between described territory;

Project described boundary node to plane and the curve of projected boundary node is passed through in matching; And

Border between the territory of reservoir model that the direction Projective Curve thus define again of the orthogonal curve of institute's matching generates.

Generated reservoir model subregion is wherein that multiple territory comprises by 14. methods according to claim 1:

Compare the parallel performance of generated subregion reservoir model subregion and the performance of the history set of the reservoir model of subregion; And

If the performance of new subregion is not equally good with the performance of history set, then rezone described reservoir model.

15. 1 kinds of methods simulating reservoir model, it comprises:

Be multiple territory by the reservoir model subregion of described generation;

The operating procedure that the simulation dividing described reservoir model comprises is multiple processing element;

Multiple processing element described in parallel processing; And

Need by balancing the load again in response to determining to simulate assessing the cost of described reservoir model, be other multiple territory by generated reservoir model subregion at least one times in described parallel process;

Analog result according to described reservoir model controls production of hydrocarbons activity;

Be wherein that multiple territory comprises by generated reservoir model subregion:

A) by select partition scheme and determine described partition scheme parameter pretreatment described in reservoir model;

B) partition scheme is used to be multiple territory by generated reservoir model subregion;

C) post processing subregion reservoir model thus correct the reservoir model of subregion;

D) quality of the subregion reservoir model of post processing is assessed; And

If e) quality of the subregion reservoir model of post processing is less than predetermined value, then repeat a, b, c, d and e with the parameter of the partition scheme of amendment and/or described partition scheme.

16. 1 kinds of methods simulating reservoir model, it comprises:

Be more than first territory by generated reservoir model subregion;

The operating procedure that the simulation being divided in described reservoir model comprises is multiple processing element;

Multiple processing element described in parallel processing; And

Need by balancing the load again in response to determining to simulate assessing the cost of described reservoir model, be more than second territory by generated reservoir model subregion at least one times in described parallel process;

Described generated reservoir model subregion is wherein that more than second territory comprises by 17. methods according to claim 16:

Identify the block of the node be isolated from each other;

Select identified joint block by size;

The joint block of weighting is selected based on processing cost; And

Distribute the joint block of weighting to corresponding territory.

Described generated reservoir model subregion is wherein that more than second territory comprises by 18. methods according to claim 16:

Determine and the processing cost level that each node in generated reservoir model associates;

The calculating weight factor of the node selected is added;

Described generated reservoir model subregion is wherein that more than second territory comprises by 19. methods according to claim 16:

Determine and the velocity field that generated reservoir model associates;

Follow the tracks of the streamline associated with described velocity field;

Project described streamline thus generate streamline curtain; And

Described generated reservoir model subregion is wherein that more than second territory comprises by 20. methods according to claim 16:

Be multiple territory according to the processing cost determined by generated reservoir model subregion.